Why this exists
The sports-pick services we’d looked at all had one thing in common: they showed off their hot weeks and quietly stopped talking when things went sideways. Win-rate claims that didn’t match the math. ROI numbers computed at -110 odds when the actual prices were materially worse. Picks that disappeared after they lost. Track records that started over every season.
So we built one we’d trust. Every pick is logged at fire time and graded after the game. ROI is computed from real captured prices wherever possible, with a clear fallback rule when we don’t have enough data. The full method is on the methodology page, including the regime-shift disclosure that other services would have buried.
Who built this
SportsGods is built by a small, quantitative product team with direct experience operating models in production. The proprietary modeling and data pipeline run server-side; public pages expose released picks, aggregate results, and user-facing context without shipping model features, vendor mappings, or decision logic.
This is a small team today. We’re honest about that because:
- Some pick services pretend to have a Wall Street trading desk. We don’t.
- The advantage of being small is we can ship fixes the day we notice the problem — and we have no incentive to hide bad weeks because there’s no quarterly investor call.
- Email hello@sportsgods.com reaches the team that built this, not an outsourced support queue.
What is written by AI
Some of what you read on SportsGods is written by a language model, and we would rather tell you that up front than let “our analyst” sound like a person at a desk. Game Report write-ups, the Daily Model Deep-Dive, and every AI Chat reply are generated by AI, automatically, with no human reviewing them before you see them.
The numbers are not. Odds, lines, public-money splits, sharp signals, model probabilities, win rates, ROI and every graded result come from statistical models and data pipelines. Grading is mechanical. The page you are reading now was written by a person.
A language model can be confidently wrong and can invent statistics, so treat AI-written analysis as a starting point and check anything you plan to act on. Methodology section 10 sets out exactly which surfaces are AI-generated and which are not.
What we promise
- Every pick on the public record. If a pick was ever sent to a subscriber, it is counted in our published record permanently — win or lose. No retroactive deletions. No selectively-reported windows.
- Real-odds ROI where we have the data, with sample-size disclosure on every aggregate stat.
- Public disclosure when something breaks — including when it makes us look worse. Our NBA team-bet models went 61.8% in the regular season and 42.5% once the 2026 playoffs started. We stopped issuing those picks, and the losing ones stayed in the record — they are still counted in the published track record. Our pause-and-reactivation policy is on the methodology page.
- No “guaranteed winner” copy, ever. Past performance does not predict future results, and we’ll be the first to tell you.
What we don’t promise
- That every signal will keep working forever. Markets adapt. Edges decay. We monitor for it and pause when the math says to.
- That you’ll be profitable. Even a +5% ROI service can have multi-week drawdowns; bankroll management and bet sizing are on you.
- Responses faster than 24 hours on email. See above — small team.
One more thing. If you ever look at a published aggregate and can’t reconcile it with the stated accounting rules, email us. We’ll explain the basis without disclosing private user data or proprietary model internals.
Responsible gambling
Sports wagering can be addictive and can cause significant harm. If you or someone you know is struggling, the National Council on Problem Gambling runs a free 24/7 helpline: 1-800-522-4700. We will refund any subscription on request, no questions asked, if responsible-gambling concerns are the reason for cancellation.